solr / lucene /弹性搜索中的文档合并中，排名/分数如何工作？

Question

I have 4 indexed documents. 1. Product category, 2. Product, 3. Product details 4. Sellers. I need to perform keyword search across all the fields of all four document and return top ranked products.

Example Product category: Electronics, Cloths and etc

Example Product : TV, Mobile in category Electronics

Example Product details: 14 in size, black colour for brand samsung TV

Example seller: Dealer 1, Dealer 2.

Example search query: "samsung LED TV"

Relationships across these documents are maintained with foreign keys. Ex: Product ducument will have category primary key.

I need a keyword search which should match against all the fields of these 4 documents and return products based on rank.

I would approach this problem by searching from child, say search seller fields, product details, product and category separately and finally join the results. The problem is ranking. Some of the more relevant products for the user keyword goes down in the result.

Is there any best way to join documents and search on all the fields of joined documents for some keywords so that the result will be more relevant and ordered?

Answer 1

The best way to join documents is before you add them to the index. Lucene is not a relational database. There are packages and such that provide similar functionality, but you'll generally be better served by providing all the content you wish to search for a document on within the document itself. I believe you want to search products here, so you could create a product Document with fields like:

Name:TV
Category:Electronics Entertainment
Seller:Dealer1 Dealer2
Details:14 in size black etc etc etc
text:TV Electronics Entertainment Dealer1 Dealer2 14 in size black etc etc etc

The text field above is added to provide a field to join everything in the document together, to easily search for anything attached to the document, which seems it would be helpful in your case, since you could then just set text as your default field when searching, or run a query like text:(samsung LED TV) .